Python BalanceTask Examples

Example #1

File: compare_with_c_source.py Project: chrisdembia/agent-bicycle

from pybrain.rl.agents import LearningAgent
from pybrain.rl.learners.valuebased import NFQ, ActionValueNetwork
from pybrain.rl.experiments import EpisodicExperiment

from environment import Environment
from tasks import BalanceTask
from training import NFQTraining

task = BalanceTask()

task.performAction(1)


print task.getObservation()

task.performAction(2)

print task.getObservation()

task.performAction(3)

print task.getObservation()

Example #2

File: balance_nfq_onlysteer.py Project: shehroze37-zz/agent-bicycle

from pybrain.rl.agents import LearningAgent
from pybrain.rl.learners.valuebased import NFQ, ActionValueNetwork
from pybrain.rl.experiments import EpisodicExperiment

from environment import Environment
from tasks import BalanceTask
from training import NFQTraining

task = BalanceTask(only_steer=True)
action_value_function = ActionValueNetwork(task.outdim, task.nactions,
        name='BalanceNFQActionValueNetwork')
learner = NFQ()
#learner.gamma = 0.99
#learner.explorer.epsilon = 0.5
task.discount = learner.gamma
agent = LearningAgent(action_value_function, learner)
performance_agent = LearningAgent(action_value_function, None)
experiment = EpisodicExperiment(task, agent)

tr = NFQTraining('balance_nfq', experiment, performance_agent)

tr.train(7000, performance_interval=1, n_performance_episodes=1, plot_action_history=True)

Example #3

from pybrain.rl.agents import LearningAgent
from pybrain.rl.learners.valuebased import NFQ, ActionValueNetwork
from pybrain.rl.experiments import EpisodicExperiment

from environment import Environment
from tasks import BalanceTask
from training import NFQTraining

task = BalanceTask()
action_value_function = ActionValueNetwork(task.outdim, task.nactions,
        name='BalanceNFQActionValueNetwork')
learner = NFQ()
learner.gamma = 0.9999
learner.explorer.epsilon = 0.9
task.discount = learner.gamma
agent = LearningAgent(action_value_function, learner)
performance_agent = LearningAgent(action_value_function, None)
experiment = EpisodicExperiment(task, agent)

tr = NFQTraining('balance_nfq', experiment, performance_agent)

tr.train(7000, performance_interval=1, n_performance_episodes=1, plotsave_interval=10, plot_action_history=True)

Example #4

File: compare_with_c_source.py Project: shehroze37-zz/agent-bicycle

from pybrain.rl.agents import LearningAgent
from pybrain.rl.learners.valuebased import NFQ, ActionValueNetwork
from pybrain.rl.experiments import EpisodicExperiment

from environment import Environment
from tasks import BalanceTask
from training import NFQTraining

task = BalanceTask()

task.performAction(1)

print task.getObservation()

task.performAction(2)

print task.getObservation()

task.performAction(3)

print task.getObservation()